RE: comment: WD 10 May 2005

Hello John

Without pre-empting more authoritative answers from SKOS editors, who will correct me as
necessary ...

> It would be helpful to have an understanding about why instances of 'Concept'
> are not instances of owl:Class, rather than owl:Thing (maybe it is and I missed
> it?).

The way I understand it, concepts in SKOS concept schemes are not generally intended to be
used as classes, but more like terms in controlled vocabularies, descriptors or subject
headings, their main function will be to support indexing of resources, and search of such
resources. Although some people will speak of concept schemes as *classification* schemes,
relationships defined by SKOS vocabulary between the indexed/classified resources and
concepts (as per skos:subject) are *not* class-instance relationships. If you want classes
and instances, don't use SKOS, use OWL :))

Confusion in this domain, IMO, comes from a wide-spread abuse of the word "taxonomy" to
refer in fact to concept schemes, although in its original meaning (e.g., in Biology),
taxonomy means a hierarchy of classes and subclasses.

> I also don't understand why a table of contents would be written in SKOS
> when it seems that a TOC contains a list of named topics/headings in the
> document, not concepts.

First I'm not sure from where you take the idea that a ToC could be expressed in SKOS. Not
in the specification, right? Unless you mean here by ToC some document *index*?. A
structured index, with maybe a hierarchy inherited from the used concept scheme, certainly
can be expressed in SKOS. But a ToC, AFAIK, is more like a structural schema of the
document, defining its sections and sub-sections (chapters, paragraphs ...). Of course,
any section defined in a ToC can be indexed as a proper resource, but the structure itself
defined by a ToC is generally orthogonal to the structure of the concept scheme used for
indexing. For example the hierarchy of sections is not necessarily consistent with the
hierarchy of concepts in the indexing scheme. Typically, sections of a document can be
re-organized for a new publication, without changing the section subjects ...

> In this vein, I couldn't find info about how SKOS
> relates to topic map integration. Practically all the examples in the
> Specification and Guide of a 'concept' are identified by plural nouns (qualified
> or not), normally all of which are indicative of a named topic.

I think there are two issues here. Practices concerning the naming of concepts can be
different from one business area to another, and from one language to another. For example
in French, a Thesaurus descriptor should be labeled as a singular noun or noun groups
(groupe nominal), with capitalization of the first letter only, for instance:

- Restauration
	-- Restauration gastronomique
	-- Restauration rapide

While some English equivalent concept schem might look more like a "taxonomy" allowing use
of plural nouns

- Restaurants
	-- Gastronomic Restaurants
	-- Fast Food

I think the intention of SKOS is to be flexible enough to cope with those various labeling
practices.

Concerning integration with topic maps, seems to me that an instance of skos:Concept will
typically be used in topic maps to identify a subject/topic (IOW as a subject indicator).
How properties of a concept will be represented in topic maps is another story. Do we need
a *canonical* representation of SKOS vocabulary in XTM, for example, I'm not sure.

> What would be disjoint with Concept, leading to the question of its scope. (Is
> it perhaps "Thing"?)

Not sure what you mean by "scope" here. But seems to me SKOS is and shoud keep completely
agnostic on what "is" and "is not" (or should be, or shoud not be) a concept (as agnostic
as topic maps on what a *subject* is, and RDF on what a *resource is*).

> The Word Net definition of 'concept' is an "abstract or
> general idea inferred or derived from specific instances"
> -- this sounds like a definition of a Class.

Sure, concepts are generally derived from specific instances, or I would say "situations".
For example the concept "Freedom of speech" is derived from situations where people are
given, or denied, the right to say what they want. Nevertheless, I am quite reluctant to
consider this concept as a class, because I wonder what the instances would be ...

> The Word Net definition of 'class' is "a collection of
> things sharing a common attribute"... So, do you see distinct properties
> associated with 'Concept' and its disjoint, and if not, what is being gained by
> using it as a flag of some sort...

The properties associated with skos:Concept are those who are defined in the SKOS
specification, therefore a skos:Concept can be anything which is declared as such and
conforms to this specification. All the rest is interpretation :))

> last, the definition of 'topic' is

> 	(1) subject, topic, theme -- (the subject matter of a conversation or
> 	discussion;
> 	"he didn't want to discuss that subject"; "it was a very sensitive topic"; "his
> 	letters were always on the theme of love")
>
> 	(2) topic, subject, issue, matter -- (some situation or event that is thought
> 	about;
> 	"he kept drifting off the topic"; "he had been thinking about the subject for
> 	several years"; "it is a matter for the police")
>
> It's not clear from the document what the relationships are between thing,
> class, concept, topic, and subject (i.e., your working data model). Any light
> you shed on this is much appreciated.

"Thing" and "Class" are defined by OWL (owl:Thing and owl:Class) and it would not make
sense for SKOS to re-define them. Since SKOS is an RDF vocabulary, and OWL-Full equals
RDF, therefore skos:Concept is a subclass of owl:Thing ... "Everything is a Thing".

"Topic" and "Subject" are defined by various Topic Maps specifications. There again, SKOS
would not be wise to re-define them.

Therefore SKOS concepts and concept schemes can be integrated on one hand in OWL/RDF
frameworks, and on the other hand can be used as subject indicators by Topic Maps
applications.

Hope that clarifies ...

----------------------------------
Bernard Vatant
Mondeca Knowledge Engineering
bernard.vatant@mondeca.com
(+33) 0871 488 459

http://www.mondeca.com
http://universimmedia.blogspot.com
----------------------------------

Received on Tuesday, 19 July 2005 09:55:35 UTC